Spoofed Video Detection Using Histogram of Oriented Gradients

Abstract

Nowadays face recognition system usage is increasing day by day to provide better security mechanism. But with the face recognition system, there are some spoofing methods also attached using which the system can be befooled easily. These attacks are simple and easy as they cost less and the images can also be easily retrieved from social sites; therefore there are high chances of them to be successful. Still there is a scarcity of a productive anti-spoofing algorithm to resolve this issue. The aim of this paper is to present a method which can be used for identification of the spoof. A method is proposed which takes the live video streaming input from the user and perform Liveness detection on the user based on the eye blinking movement and for face feature extraction Histogram of Oriented Gradient (HOG) is used as it proves to be an effective feature descriptor in the face recognition. Two classifiers k Nearest Neighbour (kNN) and Neural Network (NN) are used for the classification purpose. The work is performed on the self created Database and implementation is performed in MATLAB for better understanding, visualization and programming. The performance results of the kNN and NN classifier are compared and finally it is concluded that which classifier outperforms the other one.